After one of my colleagues posted a comment on my article, I posted back and the debate eventually devolved into the title: “How Much Do Football Stats Matter?” I said they mattered and, along with scouting, should be used in determining the worth of a player. My colleague disagreed, arguing that scouting was the only way a player could be judged. After some thought, my position has switched closer to the latter argument, but honestly, I don’t know. Here is my attempt to answer the question. This won’t be limited to one paper. Those of you interested in the other point of view, look for my paper next week.
Most of the reason for the change of my
position is the lack of usable stats at all. Unlike baseball, few stats exist in football, and what stats do exist almost entirely deal with factors that only other players can control. Completion percentage depends on the receiver catching the football and the offensive line protecting. At least seven players contribute to any completed pass. The unfairness of the statistic is that only one player-the quarterback- is actually damaged by a player failing. (Granted, a receiver could get credited with a drop, but 1.) no one thinks about drops (they rarely are the chief factor in a game), and 2.) whether or not an incompleted pass is counted as a drop or not is completely subjective. The league leader in drops had 13. How many were difficult catches that went through his hands is likely much higher.)
That raises an important question-how do you single out the quarterback? For that matter, how do you single out any position on the field? Baseball has FIP for pitchers, and a number of tools for hitters (to my knowledge there is not one single stat. WAR comes close, but also includes defense. In my mind, I use a mixture of BABIP and OPS. Both have their flaws, but BABIP helps show “luck” and contact ability while OPS takes into account plate discipline, pure power and contact ability as well. Not surprisingly, the best hitters like Miguel Cabrera do quite well. Mariners teammates Brendan Ryan and Justin Smoak not so much. For a look at all factors in a players’ game, WAR does a good job.)
To my knowledge, there is no such stat for football. Football Outsiders attempts to, but they can only compare versus the average. They do not control for the ten other players on the field. Expecting Demarco Murray to do an average job with the current Cowboys offensive line is ridiculous. The four players not named Tyron Smith in front of him simply block poorly. (Honestly, I didn’t watch another team all of last year where a first and goal at the one didn’t feel like a certain touchdown. I actually crossed my fingers in such scenarios for the Cowboys.) Tony Romo threw 19 interceptions last year, but how many were because of his horrific offensive line? The fact that he is above average despite his horrific offensive line shows that he’s good, but how much better than an average quarterback with the same offensive line? No statistic answers that question. That is because stats compare with the average player on a hypothetical team . The stats don’t take other players on real teams into account. An average player with a poor offensive gets you Felix Jones (at least I would consider him average.)
Football Outsider actually admits this point blank. They hold no punches about this fact and admit there is no way to control for the other 10 players.
Defensive stats have other problems that don’t fit neatly in this point or the next. If a cornerback makes very few tackles, is that good or bad? Is it good that he shows toughness against the ball carrier or bad that he allowed them to make the catch in the first place? Who is to say? This is touched on next, but inherent flaws in defensive stats mean that the answer is completely subjective.
Offensive football statistics tend to have another main issue- they are usage based. I call this the Larry Johnson effect. In 2006, Larry Johnson made the All-Pro team and was second the league in yards. He only accomplished those feats because he ran 416 times. LaDainian Tomlinson, the league leader in rushing and also a fellow All-Pro, came in second with 348 rushes. While Tomlinson placed fourth in yards per rush (in other words, he was efficient with his rushes), Johnson placed 15th. In that regard, Johnson was the worst All-Pro and Pro Bowler.
That raises the question: why did he get All-Pro when Maurice Jones-Drew lead all running backs in the category and didn’t get Pro-anything? The reason is that the “sexy” offensive stats are almost entirely usage based. Passing Yards, Completions, , Rushing Yards and are almost entirely based on attempts.
Matt Stafford almost threw for 5000 yards. So did Tony Romo. Normally, you and I think of 5000 yard passers as unbelievable passing phenoms like Dan Marino. Both of them are very good quarterbacks in their own right, but neither touches Dan Marino or Drew Brees. However, one place where they do touch Marino and Brees is in total attempts. Stafford, despite his 59.8% completion percentage, led the league in attempts. In fact, he broke the record. Romo placed third.
The phenomenon of yards-attempts correlation isn’t always true (Tom Brady is extremely good in this regard) but it usually is. That makes sense. Even if I am a terrible quarterback, if I throw 1000 times a season, a few of them have to be completed, right?
As discussed with Larry Johnson, the same problem presents itself in running statistics. Since 2002, only running back has lead the league in yards without placing in the top three in total attempts- Clinton Portis, in 2007. Most of the running backs are good in their own right, but never would have sniffed the top if not for touching the rock again and again and again.
Defense has its own set of problems. Defensive stats are almost entirely based on what the other team, and some statistics don’t reflect defensive realities. I think a list would organize the problems with mainstream defensive stats best.
Tackles: While also based on instincts, tackles remain a right-time-right-place proposition. Safeties can’t make tackles if they play “centerfield” against a running based team. Linebackers can’t make tackles on out routes. Defensive lineman can’t make tackles on passing plays of more than 5 yards in the air. Certain types of plays simply negate the possibility of a certain player getting a tackle, no matter the skill or instincts of said player.
Interceptions: The defensive back has to be targeted. In his prime with the Raiders, Nnamdi Asomugha only recorded more than on interception once. He has only accomplished the feat twice in his ten year career, despite consistently earning recognition as one of the leagues best shut down cornerbacks. If the receiver isn’t open, he will not be targeted. If he is not targeted, the cornerback can’t get the interception.
Forced Fumbles: While dependent on the ball carrier, there isn’t much inherently wrong with the statistic, other than the fact that there are so few forced fumbles that it the sheer quantity doesn’t always equal how hard the defensive player hits. This isn’t entirely fair, but it makes my point: Asante Samuel, probably the worst tackler in the league, had more forced fumbles in 2004 than Sean Lee has in his entire career. Again, small sample size and everything, but it does show how random the stat can be.
Fumble Recoveries: Entirely right-time-right-place. This should never be used in evaluating players. Ever. Luckily, to my knowledge, it isn’t.
Sacks: Not much wrong with this either. The best pass rushers usually get the most sacks. More pressure equals more sacks. Honestly, this seems to me to be the most fair of any of the major defensive statistics.
The numerous problems with mainstream stats show why they can never be trusted. The inherent flaws in more sabermetrically-inclined ones do too. If the sabermetrics can’t fairly discern who is actually the best and who is not, than it appears no statistic can.
The next part of this series which says that football stats can and should be used appears next weeks. I will not have a cumulative essay. You can make the decision yourself.